[0001] This invention relates to sensorless rotor position detection in reluctance machines,
particularly in switched reluctance machines.
[0002] The control and operation of switched reluctance machines generally are described
in the paper "The Characteristics, Design and Applications of Switched Reluctance
Motors and Drives" by J M Stephenson and R J Blake delivered at the PCIM'93 Conference
and Exhibition held in Nurnberg, Germany, 21-24 June 1993, which is incorporated herein
by reference. In that paper the "chopping" and "single-pulse" modes of energisation
of switched reluctance machines are described for operation of the machine at low
and high speeds, respectively.
[0003] A typical prior art drive is shown schematically in Figure 1. This includes a DC
power supply 11 that can be either a battery or rectified and filtered AC mains. The
DC voltage provided by the power supply 11 is switched across phase windings 16 of
the motor 12 by a power converter 13 under the control of the electronic control unit
14. One of the many known converter topologies is shown in Figure 2, where the power
supply 11 is provided on supply rails 26,27 which have a capacitor 25 connected across
them to cater for any alternating components of current. The phase winding 16 is connected
to the supply rails by an upper switch 21 and a lower switch 22. Energy return diodes
23,24 are connected in conventional fashion. A resistor 28 is connected in series
with the lower switch 22 to provide a current feedback signal. A multiphase system
typically uses several of the "phase legs" of Figure 2 connected in parallel to energise
the phases of the electrical machine.
[0004] The performance of a switched reluctance machine depends, in part, on the accurate
timing of phase energisation with respect to rotor position. Detection of rotor position
is conventionally achieved by using a transducer 15, shown schematically in Figure
1, such as a rotating toothed disk mounted on the machine rotor, which co-operates
with an optical, magnetic or other sensor mounted on the stator. A signal, e.g. a
pulse train, indicative of rotor position relative to the stator is generated by the
sensor and supplied to control circuitry, allowing accurate phase energisation. This
system is simple and works well in many applications. However, the rotor position
transducer increases the overall cost of assembly. It also adds extra electrical connections
to the machine and is, therefore, a potential source of unreliability.
[0005] Various methods for dispensing with the rotor position transducer have been proposed.
Several of these are reviewed in "Sensorless Methods for Determining the Rotor Position
of Switched Reluctance Motors" by W F Ray and I H Al-Bahadly, published in the Proceedings
of The European Power Electronics Conference, Brighton, UK, 13-16 Sep 1993, Vol. 6,
pp 7-13 which is incorporated herein by reference.
[0006] Many of these methods proposed for rotor position estimation use the measurement
of phase flux-linkage (i.e. the integral of applied voltage with respect to time)
and current in one or more phases. Position is calculated using knowledge of the variation
in inductance of the machine as a function of angle and current. This characteristic
can be stored as a flux-linkage/angle/current table and is depicted graphically in
Figure 3. The storage of this data is a disadvantage as it involves the use of a large
memory array and/or additional system overheads for interpolation of data between
stored points.
[0007] Some methods make use of this data at low speeds where "chopping" current control
is the dominant control strategy for varying the developed torque. Chopping control
is illustrated graphically in Figure 4(a) in which the current and inductance waveforms
are shown over a phase inductance period. (Note that the variation of inductance is
depicted in idealised form.) These methods usually employ diagnostic pulses in non-torque-productive
phases. A method suited to low-speed operation is that proposed by N M Mvungi and
J M Stephenson in "Accurate Sensorless Rotor Position Detection in an S R Motor",
published in Proceedings of the European Power Electronics Conference, Firenze, Italy,
1991, Vol.1, pp 390-393, incorporated herein by reference.
[0008] Other methods operate in the "single-pulse" mode of energisation at higher speeds.
This mode is illustrated in Figure 4(b) in which the current and inductance waveforms
are shown over a phase inductance period. These methods monitor the operating voltages
and currents of an active phase without interfering with normal operation. A typical
higher speed method is described in International Patent Application WO 91/02401,
incorporated herein by reference.
[0009] Having to store a two-dimensional array of machine data in order to operate without
a position sensor is an obvious disadvantage. Alternative methods have been proposed,
which avoid the need for the majority of angularly referenced information and instead
store data at one angle only. One such method is described in European Patent Application
EP-A-0573198 (Ray), incorporated herein by reference. This method aims to sense the
phase flux-linkage and current at a predefined angle by adjusting the diagnostic point
in accordance with the calculated deviation away from the desired point. Flux-linkage
is estimated by integrating (with respect to time) the measurement of the voltage
applied to the phase. Two one-dimensional tables are stored in the preferred embodiment,
one of flux-linkage versus current at a referenced rotor angle and another of the
differential of flux-linkage with respect to rotor angle versus current. By monitoring
phase voltage and current, the deviation away from a predicted angle can be assessed,
with the aid of the look-up tables, and system operation can be adjusted accordingly.
This method has been shown to be reliable, provided that the flux-linkage can be determined
with sufficient accuracy whenever required by the position-detecting algorithm. To
avoid the flux-linkage integrator drifting (due to unwanted noise in the system and
imperfections in the integrator) it is set to zero at the end of each conduction cycle,
when the current has fallen to zero and the phase winding is no longer linking any
flux. This method is a "predictor/corrector" method, in that it initially predicts
when the rotor will be at a reference position, measures parameters of the machine
when it believes the reference position has been reached, and uses the results of
these measurements to detect error in the prediction and hence take corrective action
by adopting a new prediction for the next reference position.
[0010] A special mode of operation of switched reluctance machines is the continuous current
mode, as disclosed in US 5469039 (Ray) and incorporated herein by reference. In this
mode, the winding is re-connected to the supply before the flux, and hence the current,
have returned to zero at the end of the energy return period. The phase windings therefore
operate with current continuously flowing through them and are always linked by flux.
This is an important mode for systems which have to produce high levels of overload
output at some points of their operating cycle. Although the efficiency of the drive
falls in this mode, it allows specifications to be achieved which would otherwise
require a larger machine. However, in this mode it has not hitherto been possible
to use any of the prior art methods for sensorless rotor position detection, as there
is no opportunity in the phase cycle to reset the integrators at some known point
of zero flux and current, since such a point does not exist.
[0011] Attempts to find a solution to this problem have included schemes which allow the
drive to operate in the continuous current mode except when the control system judges
it essential to re-estimate the position, at which time the continuous current mode
is exited, the position estimated, and the drive put back into continuous current
mode. Specifically, this can be done by running the machine in a mode which is predominantly
continuous current but drops back into discontinuous current at predetermined intervals
to allow positional information to be gained. The technique depends on the speed being
virtually constant, which may be approximately true at higher speeds (at which continuous
current is usually employed). Nevertheless, a loss of torque is associated with dropping
out of continuous current. An alternative method is to operate each phase in continuous
current for a given number of cycles, say 10, and then to excite the phase for a shorter
time on the next cycle such that the current will definitely fall to zero, allowing
the integrator to be reset and an accurate estimate of flux-linkage to be made. By
interleaving this "short" cycle with the other phases operating in continuous current,
the deleterious effect of the loss of torque is mitigated.
[0012] However, with all of these methods the loss of torque can render the machine performance
unstable and several cycles are required before stability is reached again because
the current must be built up over a period in the continuous current.
[0013] The present invention is defined in the accompanying independent claims. Some preferred
features are recited in the dependent claims.
[0014] According to an embodiment of the present invention there is provided a method of
detecting rotor position in a reluctance machine, comprising: deriving a value for
the flux linkage associated with the or at least one phase of the machine at a first
point, at a moment at which voltage is applied to that phase; deriving a value of
the phase current and the phase flux linkage at a subsequent point of the rotor; combining
the derived flux linkage values to give a value of total flux linkage at the subsequent
point; and deriving the rotor position from the phase current and the value of the
total flux linkage.
[0015] Preferably, there is provided a method in which the moment when voltage is applied
to the phase is at the point when flux-linkage growth is initiated. The current at
the said moment may be substantially zero or non-zero.
[0016] Preferably, the value of the flux linkage at the moment when voltage is applied to
the phase is derived from the current at the said moment. For example, the flux linkage
at the said moment is derived from the current and stored values of inductance for
ordinates of current.
[0017] The embodiment of the invention is, therefore, particularly useful in the single-pulse
mode of operation of a switched reluctance machine. The invention can use the value
of current at the first point to derive the value of flux-linkage. When the current
is discontinuous the zero current value gives rise to a zero value of flux-linkage.
When the current is continuous the value of current can be used to derive the non-zero
flux-linkage.
[0018] Preferably, the flux-linkage from the said first point is derived by integrating
the phase voltage from the said moment to the subsequent point. The said flux-linkage
at the subsequent point may be derived by integrating the phase voltage from the said
moment to the subsequent point. The rotor position may be derived from stored parameters
having coordinates of phase current and flux-linkage.
[0019] The method of the invention according to one particular embodiment measures the current
at turn-on of a phase winding when flux growth is initiated and uses this current
value to index a table of inductance. The value of inductance provided by the table
is then multiplied by the current to give an estimate of the standing flux-linkage
in the phase. Also at turn-on, a flux-measuring integrator, which is set to zero,
is put into integration mode. At a predetermined subsequent point, the value of flux-linkage
provided by the integrator is added to the calculated value for standing flux-linkage,
and the resulting total is used to determine rotor position.
[0020] Because the inductance is determined at switch turn-on, i.e. the initiation of flux
growth in a phase inductance cycle, the method does not require large amounts of stored
data. It is also robust in the presence of noise on the waveforms from which it deduces
position.
[0021] Also according to the present invention there is provided a method in which rotor
position is derived from values associated with each phase of a polyphase machine.
[0022] One embodiment of the present invention provides a robust and cost-effective method
of monitoring rotor position and a rotor position detector, which can operate without
a rotor position transducer in the single-pulse mode, with or without continuous current.
[0023] The invention can be put into practice in a number of ways, some of which will now
be described by way of example and with reference to the accompanying drawings in
which:
Figure 1 shows a typical prior art switched reluctance drive;
Figure 2 shows a known topology of one phase of the converter of Figure 1;
Figure 3 shows typical flux-linkage and phase current curves, with rotor position
as a parameter;
Figure 4(a) shows a typical motoring current waveform in chopping control;
Figure 4(b) shows a typical motoring current waveform in single-pulse control;
Figure 5 shows in schematic form a switched reluctance drive in which the invention
is embodied;
Figure 6 shows flux-linkage waveforms in the drive of Figure 5 in continuous current
mode;
Figure 7 shows a continuous current waveform for the drive of Figure 5; and
Figure 8 shows flux-linkage waveforms in the drive of Figure 5 in discontinuous current
mode.
[0024] The phase inductance cycle of a switched reluctance machine is the period of the
variation of inductance for the, or each, phase, for example between maxima when the
stator poles and the relevant respective rotor poles are fully aligned. The illustrative
embodiment to be described uses a 2-phase switched reluctance drive in the motoring
mode, but any phase number from one upwards could be used, with the drive in either
motoring or generating mode.
[0025] Figure 5 shows a system for implementing the method in which the invention is embodied.
Figure 7 illustrates graphically the continuous current waveform for the system of
Figure 5. In this system, a power converter 13 is typically the same as that shown
in Figure 1, and like reference numerals have been used where appropriate. The converter
13 controls the switched reluctance machine, as before. The converter 13 is itself
controlled by a controller 42 which, in this embodiment, incorporates a digital signal
processor 44, e.g. one from the Analog Devices 2181 family. Alternative embodiments
could incorporate a microprocessor or other form of programmable device, as are well-known
in the art. The illustrated 2-phase machine has a stator 30 and a rotor 32. The stator
has four poles 50, on which are wound phase windings 34/36. The rotor has rotor poles
52 and, to assist with starting the machine, has a pole face 54 that defines a stepped
airgap with the face of a stator pole. One skilled in the art will realise that a
machine with a different phase number or pole combinations could be used, since the
invention is not specific to any particular machine topology. Similarly, the invention
is not restricted to a particular type of control technique and any controller and
converter can be used as long as it is suitably programmable.
[0026] Phase current is sensed by a current transducer 38 arranged in relation to each phase
winding. The output signals indicative of current in each phase are fed to the controller
42. A look-up table 46 storing phase inductance for rotor angles is also connected
with the controller 42. While a current transducer for each phase is shown, one or
a selection of phases could be monitored for phase current according to the invention.
[0027] An integrator 40 depicted in the controller 42 is used to derive measurements of
flux by integrating the phase voltage V across the winding provided by voltage transducer
43. For greater accuracy the voltage drop (IR) across the winding can be factored
out of the integrated value. Only one voltage transducer 43 is shown in Figure 5 as
the supply voltage will be common to each phase. It will be appreciated that each
phase could have a respective voltage transducer. While the integrator is shown as
a discrete device 40 it is preferably implemented in the software running in the processor
44.
[0028] The method according to this embodiment of the invention operates as follows. It
is assumed that the machine is operating in continuous current mode and that the rotor
position is known sufficiently well to allow the winding to be energised. At the point
of turn on, when positive voltage is applied across the phase, the value of current
is measured by the transducer 38 and held by the controller 42. Knowing the turn-on
angle, the look-up table 46 of phase inductance against angle is interrogated to return
the phase inductance corresponding to the rotor angle. The value of inductance is
multiplied by the stored current value to give the flux-linkage in the phase at the
point of turn on. This value is stored. As the rotor turns, the integrator 40 operates
to integrate the voltage across the phase winding. When the control system determines
that the subsequent predetermined position has been estimated to have been reached,
the current is measured using the transducer 38 for the active phase and the estimate
of flux linkage from the integrator 40 is added to the stored value from the table
46 to give the total value of flux linkage in the phase at that position. This total
value is then used, in conjunction with the current, to find the true position which,
if necessary, can be used to correct the previous estimate.
[0029] The angular difference Δθ between the predicted rotor position θ
m and the reference rotor position θ
r can be calculated by the processor 42 as
[0030] In order to determine the angular difference between the predicted position (at which
the measurement of flux and current are made) and the reference position , it is therefore
also a feature of this embodiment to store in the processor for the reference position
values of partial derivative ∂θ/∂ψ (or its inverse ∂ψ/∂θ) for a set of values of phase
current i.
[0031] Since the reference rotor position θ
r is known, the true rotor position at the predicted instant in time may be calculated
from Δθ as
[0032] The expected time to the next rotor position can then be estimated using the known
value of motor speed. Under accelerating or decelerating conditions a correction may
need to be made to the motor speed.
[0033] If all phases are used for measurement the next rotor reference position will be
that for phase 2 (or whatever is the next phase in the sequence). For an n-phase motor
with p rotor poles, the angle of rotation to this position will be (360/np)° - Δθ
and, by dividing this angle by the speed, the time required to reach this next position
can be estimated.
[0034] If only one phase is used for measurement, the next rotor reference position will
be after a rotation of (360/p)° - Δθ and, by dividing this angle by the speed, the
time required to reach this position can be estimated.
[0035] The predicted time to the next reference position is then counted out using a high
frequency clock (not shown) by known means and at the instant such time has elapsed
a further measurement of flux Ψ
m and current I
m is made for the corresponding phase. Due to changes in speed, and tolerances in stored
data and calculations, the predicted position θ
m will not be identical to the reference position θ
r. The angular difference θ
r - θ
m can again be calculated using equations (1) and (2) and the procedure outlined above.
[0036] The process of predicting rotor positions on a phase by phase basis and measuring
the true rotor position at each measurement instant is repeated to provide an incremental
indication of rotor position as a direct replacement for existing optical or electromagnetic
rotor position sensors.
[0037] The procedure for the case of single-pulse operation under regenerating conditions
is the same as described above except that the values of ∂θ/∂ψ shown in Figure 6 will
be negative (rather than positive) for the corresponding typical reference rotor position.
[0038] Various arrangements may be used for the measurement of flux. Any known form of flux
transducer could be used. However, the measurement is preferably by means of analogue
or digital electronic resettable integrators (which integrate the phase voltage with
respect to time t), together with means for compensating for the resistive voltage
drop in the phase winding.
[0039] The integrator executes the equation:
where: v is the phase voltage
i is the phase current
R is the phase resistance
t
m is the instant of measurement
[0040] The integrator start time t
o is arranged to be the instant of application of voltage to the phase each time the
phase is energised (for the higher speed mode). The digital processor is informed
of the instant of application of the voltage to the phase by the electronic controller
using a control interface (not shown). The integrator is reset by the digital processor
after each flux reading has been made.
[0041] In applications for which the direct source voltage V is relatively large compared
to the voltage drop across the semiconductor switches in the power converter, the
direct source voltage may be measured and integrated in place of the individual phase
voltage. This has the advantage that only one voltage need be measured.
[0042] The flux is then measured as:
[0043] As an alternative to using separate electronic integrators, the digital processor
may be used by multiplying the direct supply (or phase) voltage by time on a step
by step basis. This has the disadvantage that the digital processor is substantially
busy and may need to be a separate unit.
[0044] However, provided the supply voltage is substantially constant and large compared
with the resistive voltage iR, various approximations may be made. For example:
where k is a constant, typically k = 0.5, such that:
[0045] As a further embodiment, in the case where the supply voltage V is relatively large
compared with the resistive voltage iR, the need to compensate for the resistive voltage
drop may be avoided by using a modified value for the flux in the stored data or ignoring
iR altogether.
[0046] In this case the values of phase flux linkage ψ for a particular current I and particular
rotor position θ stored in the digital processor are replaced by values of the volt-second
integral ψ' required to create the phase current I for the rotor position θ are given
by
[0047] In testing the machine to establish the table of values of ψ' and ∂θ/∂ψ', v may,
for convenience, be held constant (provided v is relatively large) and is preferably
equal to supply voltage. The rotor position measurement procedure is the same as already
described in this application except that the measured flux ψ
m and expected flux ψ
e and partial derivative (∂θ/∂ψ) are replaced by ψ'
m' ψ'
e and (∂θ/∂ψ') respectively, where ψ'
e and (∂θ/∂ψ') are obtained as described from the stored data represented by Figures
4 and 6 and where the flux ψ
m' is measured as
[0048] Equations (3), (4), (5) and (6) represent different methods of evaluating the phase
flux linkage for the purpose of identifying rotor position and these represent different
implementations of the technique.
[0049] The various embodiments described above are all based on the measurement of flux
ψ
m and current I
m at a predicted rotor position, the look-up of the expected flux ψ
e for the measured current i
m corresponding to the reference rotor position, and the calculation of the difference
Δθ between the reference rotor position and the predicted rotor position according
to the equation:
[0050] Figure 6 illustrates the method of the invention graphically. The saw-tooth waveform
ψ
a represents the actual flux linkage associated with a phase while the machine is in
the continuous current mode. The flux-linkage value ψ
s is the minimum, or "standing" value during the cycle. The saw-tooth waveform ψ
i represents the flux linkage indicated by the output of the integrator. The phase
is switched on at the angle θ
on, prior to which the integrator has been held in reset for the period R by means of
a signal from the processor 44. This period R is long enough to return the integrator
output to zero. At θ
on, the current is measured (see Figure 7) and the look-up table 46 of inductance is
interrogated to find the inductance of the phase for that rotor angle. The product
of the current and the inductance is calculated by the processor 44. This gives an
estimate of ψ
s, which is then stored by the controller 42. After θ
on, the actual flux linkage in the phase increases at a rate dictated by the applied
voltage, and is tracked, with an offset of ψ
s, by the output of the integrator. At some point (not critical to this discussion)
the phase is switched off and the flux linkage begins to ramp down. When the control
system determines that the rotor is at the predetermined position θ
ref, the output from the integrator and the value of phase current are sampled and held.
The stored estimate of ψ
s is added to the value of ψ
i to give an estimate of ψ
a. The current and ψ
a can then be used to find the actual rotor angle in the way described above and taught
by Ray in EP-A-0573198.
[0051] This embodiment of the invention is particularly advantageous in that it works equally
well when the phase current is discontinuous, i.e. in the conventional single-pulse
mode. This is illustrated graphically in Figure 8. Since the current is zero immediately
before θ
on, the multiplication with the inductance value gives the correct result of zero standing
flux-linkage. Thus, the same program code can be used in the controller 42 for both
discontinuous and continuous current.
[0052] The method described above provides a simple, yet effective, way of combining continuous
current operation with sensorless position detection, without any unwanted degradation
in the performance of the machine.
[0053] The skilled person will appreciate that variations of the disclosed arrangements
are possible without departing from the invention, particularly in the details of
the implementation of the algorithm in the controller. Also, the diagnosis on which
rotor position detection is based could be carried out in only one phase of a polyphase
machine. Accordingly, the above description of several embodiments is made by way
of example and not for the purposes of limitation. It will be clear to the skilled
person that minor modifications can be made to the drive circuit without significant
changes to the operation described above. The present invention is intended to be
limited only by the scope of the following claims.
1. A method of detecting rotor position in a reluctance machine, comprising:
deriving a value for the flux linkage associated with the or at least one phase of
the machine at a first point, at a moment at which voltage is applied to that phase;
deriving a value of the phase current and the phase flux linkage at a subsequent point
of the rotor;
combining the derived flux linkage values to give a value of total flux linkage at
the subsequent point; and
deriving the rotor position from the phase current and the value of the total flux
linkage.
2. A method as claimed in claim 1 in which the moment is at the point when flux-linkage
growth is initiated.
3. A method as claimed in claim 1 or 2 in which the current at the said moment is non-zero.
4. A method as claimed in claim 1 or 2 in which the current at the said moment is zero.
5. A method as claimed in any of claims 1 to 4 in which the value of the flux linkage
at the moment when positive voltage is applied to the phase is derived from the current
at the said moment.
6. A method as claimed in claim 5 in which the flux linkage at the said moment is derived
from the current and stored values of inductance.
7. A method as claimed in any of claims 1 to 6 in which the flux-linkage at the said
subsequent point is derived by integrating the phase voltage from the said moment
to the subsequent point.
8. A method as claimed in claim 7 in which the said moment is the point of minimum flux-linkage
and the integration is started at the said moment.
9. A method as claimed in any of claims 1 to 8 in which the rotor position is derived
from stored parameters of phase current and flux linkage.
10. A method as claimed in any of claims 1 to 9 in which rotor position is derived from
values associated with each phase of a polyphase machine.
11. A method as claimed in any of claims 1 to 9 in which the rotor position is derived
from values associated with one phase of a polyphase machine.
12. A rotor position detector for a reluctance machine comprising:
means for determining a value for the flux linkage associated with the or at least
one phase of the machine at a first point, at a moment when voltage is applied to
that phase;
means for deriving a value for the phase current and the phase flux linkage at a subsequent
point of the rotor;
means for combining the derived flux linkage values to give a value of total flux
linkage at the subsequent point; and
means for deriving the rotor position from the phase current and the value of total
flux linkage.
13. A detector as claimed in claim 12 including a look-up table storing values of inductance
for values of phase current.
14. A detector as claimed in claim 12 or 13 in which the means for deriving a value for
the flux linkage at the subsequent point include an integrator for integrating the
phase voltage from the said moment to the subsequent point.
15. A detector as claimed in claim 14 including reset means operable to reset the integrator
for the said moment.
16. A detector as claimed in claim 12, 13, 14 or 15 including processor means operable
to determine the value for the flux-linkage when flux-linkage growth is initiated.
17. A detector as claimed in claim 16 in which the processor means are operable to derive
the value for the flux-linkage from the current at the said moment.
18. A detector as claimed in any of claims 12 to 17 in which the means for deriving the
rotor position include storage means for storing values of rotor position for values
of phase current and flux linkage.